The question

How do you go deep without turning the work into an indefinitely growing pile of reading, branches, and experiments?

Working model

A useful research step must still be able to change a decision. Pin the exact source and the result owner, preserve the first surprising observation, and make each plausible explanation face a test that could make it lose. Separate “this mechanism can work” from “this is the smallest justified change.”

Stop when the next probe cannot change the selected mechanism, claim boundary, or reopening condition. A careful negative result is a real result when it records what lost, what remains possible, and what new evidence would justify reopening.

Why it matters

Without a discriminator, more depth mostly creates narrative momentum. With one, even a long investigation keeps shrinking the decision surface.

Provenance

Synthesized from Linux Fieldwork at revision c79d34b65fb2cd8c54234f361f073248c53b513a. The source note describes a research method; it does not claim every investigation needs the same machinery.

Take one open technical question and write four lines:

1. Decision: what choice could this work change?
2. Discriminator: what result would make the leading explanation lose?
3. Stop rule: when is the current question saturated?
4. Reopen trigger: what new evidence would make another pass worthwhile?

If line 2 is impossible to write, the question is probably still too vague.